BlackRock aims to raise $36 billion to bet on Anthropic! The "securitization" of AI computing power is skyrocketing; who will pay for the potential credit defaults?

Blackstone, the giant in private equity and alternative asset management on Wall Street, has initiated preliminary discussions with investors to gauge market interest in a second massive AI-related debt financing scheme. This funding will support AI leader Anthropic in utilizing TPU power chips from Alphabets Google. Media reports citing informed sources indicate that one of the preliminary plans aims to raise at least $36 billion in debt. These sources also stated that specific details, including the financing size, deal structure, and whether Blackstone will ultimately lead the financing, are still under discussion and may change. Due to lack of authorization to comment publicly, these individuals preferred to remain anonymous. These large-scale AI financings primarily strengthen the order certainty in the computing power supply chain, with concerns regarding the AI bubble burst and liquidity risks coming second. The second financing being discussed by Blackstone, with a minimum of $36 billion, layered on top of the $35 billion AI XPV initial deal already completed by Apollo and Blackstone, essentially translates some of Anthropic's expected future revenues from Claude into substantial current orders for Googles TPU AI computing power infrastructure, optical communication/interconnection, data center CPUs, high-performance Ethernet switching infrastructure, data center HBM/DRAM/NAND storage components, and data center power supply chains. The initial financing of the AI XPV platform corresponds to over 1GW of computing power and is planned to support over 20GW of deployment before 2028; this means that chip orders are no longer entirely dependent on the current cash flow from AI labs. Instead, they can be market capitalized through SPV, long-term leasing, and vendor credit support. Consequently, the visibility of orders for Broadcoms AI power supply chain, Googles TPU ecosystem, network equipment, and core hardware suppliers for power and data centers will be significantly extended in the short term. The risk lies in the fact that financial engineering has not eliminated repayment risk; it has merely transferred it from the AI labs balance sheet to the SPV creditors, chip suppliers, cloud computing platforms, and their guarantors. This wave of AI prosperity is not necessarily doomed to fail, but the core of its pricing is shifting from technology penetration rates to whether the credit chain can be supported by real cash flows. Castle Securities projects that by 2028, debt financing for AI chips could exceed $500 billion, making the phrase credit leverage risk is poised to explode a reasonable cautionary note for analysts. AI financing competition sets new records: Blackstone proposes to raise at least $36 billion, betting on Anthropics computing power landscape with Googles TPU chips. If this transaction ultimately approaches the stated scale, it will surpass the $35 billion in debt financing arranged by another Wall Street asset management giant, Apollo Global Management, and Blackstone for Anthropics exclusive customized AI chips developed by Google about two months ago. This financing is one of the largest credit transactions in the history of the private credit market. Representatives from Blackstone, Apollo, Anthropic, and Google all declined to comment. As Silicon Valley races to build artificial intelligence infrastructure, major leading companies have entered into complex financing deals that often have circular characteristics to secure computing resources. Google is one of Anthropics earliest investors and has repeatedly purchased its equity assets, and now increasingly provides guarantees to support the massive construction of data centers for this AI startup. This potential new round of financing follows Anthropics secret submission of an IPO application to the U.S. stock market. The company is trying to enter the public market ahead of competitor OpenAI. As the developer of the Claude AI large model series, Anthropic plans to rent high-performance computing chips in five data centers with the help of Google, supported by the previous AI debt financing arrangement. Global tech companies are tapping into various segments of the credit market to meet the unprecedented capital demands of artificial intelligence, which compels Wall Street asset management giants to collectively participate in designing new debt structures to keep pace with industry expansion. Due to market concerns that AI investments may not yield the anticipated positive returns, some companies have recently been forced to pay high benchmark yields when issuing new debt. Broadcom, Apollo, and Blackstone established a collaborative platform named AI XPV Platform this year to help leading AI technology development companies, including Anthropic, finance AI computing power infrastructure. The $35 billion AI-related debt financing completed about two months ago is the first financing provided by this platform. In this deal, Broadcom provided backstop financing support for the largest senior debt portion. Previous reports indicated that Morgan Stanley acted as Broadcoms advisor and assisted in arranging the transaction. Representatives from Broadcom did not respond to requests for comment. Wall Street begins the process of securitizing computing power: Anthropic and Meta are cashing out part of their future demand in advance, as the AI bull market enters the credit leverage stage. The $35 billion AI XPV financing that has been executed supports Anthropics expansion of over 1GW of computing power, while Blackstone is discussing a second financing with a preliminary scale of at least $36 billion. Blackstone's proposed new round of approximately $36 billion financing, layered on top of the executed $35 billion AI XPV initial funds, is converting a portion of Anthropic's strong future revenue expectations into current TPU and data center orders; this platform initially supports over 1GW of computing power and aims to drive the global deployment of over 20GW of AI infrastructure. This means that chip orders are no longer entirely dependent on the current cash flow from AI labs, but can be market capitalized by the Wall Street asset management giants through SPV, long-term leasing, and vendor credit support. Consequently, the visibility of orders for Broadcoms AI power supply chain, Googles TPU ecosystem, network equipment, and core hardware suppliers for power and data centers will be significantly extended in the short term. The risk lies in the fact that financial engineering has not eliminated repayment risk; it has merely transferred it from the AI labs balance sheet to SPV creditors, chip suppliers, cloud platforms, and their guarantors. Anthropics financing structure involves the SPV purchasing Google TPUs and leasing them to Anthropic, with the largest senior debt portion relying on Broadcom providing gap or residual value support, thus allowing around $25 billion of core debt to have a yield reduced to approximately 5.75%; subordinate debt without equivalent guarantees has a yield of about 8.5%. This creates a typical vendor financing closed loop: chip manufacturers facilitate customer procurement through credit endorsements, and the procurement orders convert into revenue and valuation bases for chip manufacturers. Should Anthropics revenue growth, computing power utilization, or IPO financing fall short of expectations, the risks may reverse along the chain of lease paymentsSPV debt repaymentchip residual valuevendor guarantees. Meta, the parent company of Facebook, shows that such off-balance sheet capital commitments have risen from project-level arrangements to being a core financing model for giant tech companies. As of the end of June 2026, unleased commitments not yet recorded as lease liabilities on the balance sheet have reached $278.99 billion, a significant expansion compared to the previous quarter; in July, Meta signed approximately $68 billion in data center lease agreements with terms of 18 to 20 years, while also having $349.31 billion in irrevocable contract commitments and potential cloud capacity procurement obligations up to $14.72 billion. Meanwhile, Meta has raised its capital expenditure guidance for 2026 to $130 billion to $145 billion. Strictly speaking, these large AI financings and debt projects are not hidden debts in the traditional sense, but their economic essence represents long-term fixed payment obligations: when facilities go live, lease liabilities will gradually enter the statements, while cash flow pressures have already been locked in through contracts. Castle Securities forecasts that by 2028, new debt issuance solely for AI chip procurement may exceed $500 billion, accounting for over 5% of the U.S. investment-grade bond index by then, with single-year issuance in 2028 potentially exceeding $250 billion, most of which will have maturity terms of only three to five years to match the shorter economic lifespan of the chips. The real danger lies not just in too much debt, but in the maturity mismatch between short-duration debt, rapidly depreciating assets, and long-term, highly uncertain AI revenues: should interest rates remain elevated, model inference prices drop, chip architectures update causing a sharp decline in the residual value of old equipment, or the credit market demand higher issuance premiums, many projects may simultaneously face a refinancing wall around 2028, squeezing allocations in technology, media, and telecommunications bonds, leading to credit spread expansion and a contraction in private credit liquidity. Therefore, some seasoned analysts on Wall Street believe that investment strategies should separate the valuation of financed orders from terminal cash demand: prioritize allocations to clients that are diversified, cash-rich, have orders with prepayment or irrevocable commitments, and can continuously generate free cash flow in the chip and infrastructure sectors; for projects relying on SPV, vendor guarantees, cyclical investments, and single AI lab rent repayment, risk discounts must be increased, and credit spreads, CDS, lease-adjusted leverage ratios, chip residual values, and computing power utilization rates should be closely monitored. Overall, this wave of AI prosperity is not necessarily destined to rupture, but the core of its pricing is shifting from technology penetration rates to the robustness of the credit chain in being supported by real cash flows.